Granular pitch-shifter / phase vocoder (via fast Fourier anaysis) delivers a wide range of dramatic contemporary effects. Whole grain goodness...
Effectively cuts audio material into snippets ('grains'), then plays them back at a variable rate – even backwards. In the original tape devices, the radius of a rotating cylindrical tape-head determined grain size and the length of tape in contact with the cylinder determined the overlap between grains. But Uhbik-G is not bound by the physical limitations of rotating cylinders – you can make it huge or impossibly small!
Pitch Shifter History
The ancestor of today’s pitch-shifting and time-stretching devices was developed in Germany by the conductor Hermann Scherchen (1891-1966), the first Apparatus for Independent Control of Pitch and Tempo of Audio Recordings (original: Apparat zur unabhängigen Kontrolle von Tonhöhe und Tempo von Tonaufnahmen). Little information is available, but development is likely to have been during the 1930s, shortly after the invention of magnetic tape recorders.
Scherchen’s device had four playback heads attached to a small rotating drum which was mounted between the original playback head and the capstan of a conventional tape recorder. When the drum is rotated, the heads take turns to read the tape, but unlike fixed playback heads this can happen at a variable rate. As long as the drum is not rotating, the pitch remains unaffected. If the drum is rotated in the opposite direction to the tape, very short (smoothly) overlapping samples of the recording are played back while gaps between them are effectively skipped, resulting in a higher pitch. The opposite effect happens when the drum is moving in the same direction as the tape - the pitch is lowered because multiple samples of the same audio material are sent to the output. The effect turns into time-stretching when the tape itself is slowed down or sped up to compensate for the pitch change.
This same principle was applied in commercial devices e.g. the Eltro Information Rate Changer used so effectively for HAL’s death scene in 2001 - A Space Odyssey. A short article about this on Wendy Carlos’ own website: http://www.wendycarlos.com/other/Eltro-1967/index.html
Although the contraption may seem bizarre, the same basic principle is applied in modern granular pitch-shifters, which appeared as hardware in the ‘80s then as software in the ‘90s. Pitch-shifters effectively cut the audio material into small overlapping snippets and play them back at a variable rate. The snippets are now called grains, and their duration is the grain size. In the old tape devices, the radius of the drum determines grain size while the length of tape in direct contact with the drum determines the overlap between grains. Uhbik-G is not bound by the physical limitations of rotating drums – you can make the ‘drum’ impossibly small!
This parameter is controlled using the grainsize knob, up to about 2 seconds. Unlike conventional pitch-shifters, it affects the duration of grains as they appear at the output. The input grain size is adjusted automatically.
There are two parameters controlling pitch:
semitone: adjusts the pitch of grains in semitones (+/- 12), and is therefore suitable for precise musical intervals.
scale: multiplies the rate of grain playback, with a range of 0 to 4 octaves. A scale value of 1 means the original pitch, values between zero and 1 lower the pitch. At zero, only a single sample is played back, but because grains still overlap, this results in an effect reminiscent of low-pass filtered sample-rate reduction. The scale knob is bipolar: negative values play grains backwards, and this can deliver some very interesting reverse effects.
Semitone and scale can be used at the same time: For instance, play grains backwards (scale = -1) while transposing them up an octave (semitone = +12).
Like most Uhbik effects, Uhbik-G includes an offset parameter which scales the effect differently per channel. For instance, if you set scale to zero and offset to 100, the grains in the right channel will play back normally while those in the left channel are played in reverse.
The mix parameter determines the relative volumes of the original (dry) and effect (wet) signals.
The reset and auto reset switches are used for synchronizing the start of grains (after a moment of silence) to suit the audio input material – either manually or automatically. This is quite subtle, but can be very useful for the timing of effects.
The iteration parameter feeds the output back into the effect input, resulting in multiple detuning e.g. echos with constantly rising or falling pitch.
Phase Vocoder Mode
Uhbik-G can be switched into a fundamentally different mode called PhaseVoc by clicking on the operation button. This mode has three quality levels instead of grain size (the knob disappears).
In PhaseVoc mode, pitch shifting is achieved by time-stretching or time-compressing the spectrum of the input signal. The signal is split into its component sine waves via Fourier analysis, and the phase and position of these waves are adjusted using the scale and semitone knobs.
In PhaseVoc mode, grains cannot be played in reverse – negative scaling is interpreted as positive.
The result is highly dependent upon the audio material used. While granular often sounds rather rough (it’s granular!), phase vocoding can often sound mushy because most of the transients are lost. However, when applied to vocals and pads, this effect can be very impressive!
Like all effects based on FFT (Fast Fourier Transformation), there is a noticeable latency between the input and output signals. In Uhbik-G, this has intentionally been left uncompensated because the granular algorithm itself would otherwise have required additional delays. If you want to use the PhaseVoc mode for rhythmically critical material (despite the loss of transients), you can move the audio track forward in time by about 2000 samples.
The Phase Vocoder mode can be extremely CPU draining. Use with caution! Set to lowest quality whenever sufficient!